This paper compares and contrasts the nature and scope of change in the domestic climate governance of India and South Africa between 2007 and 2010. It uses an actor‐centered approach to analyze the drivers of change. An exploratory test of fit shows that the concept of "communities of practice" captures the trends and actor relations well for the South African case, while more simple networks could be identified in India. Using data from an expert survey and from semi‐structured interviews, this paper finds that both countries have generally not yet surpassed the level of second‐order change, or double‐loop learning. Differences exist for more specific parts of climate governance. Three resulting hypotheses give conditions for the development of either communities of practice or of networks, as conceptualized in formal network analysis. They target (1) the number of participating actors, (2) the size of the scientific landscape and the degree of competition among scientists, and (3) the centrality of a governmental actor with a certain knowledge and attitude within a network.